Add a test for masked arrays hiding within namedtuples

[ev-br]

A minimal fix is craft a test which fails before #148 and passes after.

In general, handling of e.g. stats.mstats.describe is a bit brittle, maybe there is a structural solution.

[Sheila-nk]

Hi @ev-br
I am currently working on this issue.
stats.mstats.describe expects a masked array instead of a masked array within a namedtuple so I can see why it doesn't work well with our example.
Wondering if this is a more straightforward test:

  >>> import numpy.ma as ma
  >>> from collections import namedtuple
  >>> x = ma.array([90, 20, 35, -10, 52], mask=[0, 0, 0, 1, 0])
  >>> Data = namedtuple('Data', ['student', 'scores'])
  >>> record = Data(student='Rose', scores=x)
  >>> record
  Data(student='Rose', scores=masked_array(data=[90, 20, 35, --, 52],
           mask=[False, False, False,  True, False],
     fill_value=999999))

It does pass doctesting but please clarify if I misunderstood the issue.

[ev-br]

Great! IIRC, these sorts of tests were going through the vanilla doctest, with literal comparisons. I'd try to use floating point inputs and deliberately change the want layout.